Solar radiation plays an important role in the design and analysis of energyefficient buildings in different climates. Solar availability in China is excellent with more than two thirds of the areas having 2200 hours of sunshine and the annual solar radiation in excess of 5860 MJ/m2. In China, there are cities/regions that do not have measured solar radiation data. The primary aim of this study is to model global solar radiation (GSR) in different climate zones in China using 2 methods – regression analysis and artificial neural networks (ANNs), evaluate their performance and investigate the impact of using modelled GSR on building energy simulations. Correlation between clearness index and sunshine duration is useful in the estimation of solar radiation. Regressions and ANNs were employed to predict the daily global solar radiation in China. Measurements made during the 30-year period (1971 to 2000) from 41 measuring stations covering 9 thermal and 7 solar climate zones across China were gathered and analysed. Two-parameter Angstrom-Prescott linear regression equation was used to investigate the correlations between daily global solar radiation and sunshine duration. Three sets of models were considered - individual city models, thermal climate zone models and solar climate zone models. ANNs were also employed to generate prediction models for daily GSR using 6 input variables – day-number, latitude, longitude, altitude, dry-bulb temperature and sunshine duration. The performance of the regression and the ANN models was compared and analysed. The coefficient of determination (R2), Nash-Sutcliffe efficiency coefficient (NSEC), mean bias error (MBE) and root-mean-square error (RMSE) were determined. Regression models showed a strong correlation between the clearness index and sunshine duration in both thermal and solar climate zones (R2 = 0.79-0.88). Both the ANN and the regression models had similar NSEC (0.8-0.95), revealing a reasonably good predictive power. An increasing trend of larger MBE and RMSE from colder climates in the north to warmer climates in the south was observed. To investigate the impact of using modelled GSR on heating and cooling loads, building energy simulations were conducted for 9 cities (Harbin 45o45'N, Hami 42o49'N, Dunhuang 40o09'N, Minqin 38o38'N, Lanzhou 36o03'N, Yushu 33o01'N, Yichang 30o42'N, Lijiang 26o52'N and Hong Kong 22o18'N) in the 9 thermal climate zones with the simulation tool VisualDOE 4.1. Two weather files in DOE-2 weather file format were developed for each city, one with the measured GSR and the other predicted data from the regression thermal zone models. Weather data measured in year 2000 were used in the simulation exercise. A generic office building was developed for each city based on the local energy codes and the prevailing architectural and building construction engineering practices. Two simulations per city were carried out, with the 2 weather files developed. Annual building cooling loads (using measured GSR) ranged from 5499 MWh in Harbin to 9357 MWh in Hong Kong and heating loads ranged from 221 MWh in Hong Kong to 1794 MWh in Harbin. The MBE between the cooling loads using measured and predicted GSR ranged from -2.3% in Lijiang to 1.5% in Harbin, and the heating loads from -6.2% in Lanzhou to 1.8% in Lijiang. These suggested that the difference (i.e. modelled against measured GSR) in the computed cooling loads ranged between 2.3% underestimation and 1.5% overestimation, and heating loads between 6.2% underestimation and 1.8% overestimation. These findings could give architects and building engineers an idea about the likely variations in computed heating and cooling loads.